Predistion Nino 3.5 SST anomaly

The impact of global climate disturbance like El Nino-La Nina on human life and its environment tends to increase from year to year in the last 2 decades and may be one of the most serious problems of the next millenium. To mitigate the impact we should make an improvement on the numerical model prediction ability which have only 3 months accurateness. Using Fourier theorem, we elucidated NINO3.5 SST anomaly time series by applied Fast Fourier Transformation (FFT) and obtain 6 main periods. By trial and error, we generated harmonic model to predict NINO3.5 SST anomaly but the result is not sufficient because the periodicity of model. By applied Phase Dispersion Minimization (PDM) method, we found another period ie. 68.50 months/cycle and put it as harmonic. By added this period in the harmonic model, the periodicity is disappear. To increase the prediction accurateness we have applied the artificial neural network and trained the model. After running for 30 epoch we could improved the correlation between model and data from 0.506 to 0.812. Analysis on the harmonics showed that for El Nino event since 1950 are dominated by 3 first harmonics meanwhile for the La Nina event by the last 4 harmonics. Comparison the El Nino 82/83 to El Nino 97/98, showed that the El Nino 82/83 dominated very strong by the 1", 2" and 6" harmonics and moderate by 3 and 5 harmonics, meanwhile El Nino 97/98 dominated very strong by 3, 4, 5, and 6" harmonics and strong by 1" and 4" harmonics. These condition may could explain why El Nino 82/83 slightly stronger then El Nino 97/98. From all of harmonics, we known only about 2 of them, ie. 4 and 6" as half of sunspot and QBO period and we supposed that way to understood El Nino/La Nina is to understand these 7 harmonics.